Tag Archives: davos

Forget ‘Intel Inside’; It’s Now AI Inside

I am sure we are all familiar with the Intel slogan, “Intel inside.” This has been a very powerful tagline and one that has helped Intel become the dominant PC chip supplier. (I know I was very influenced by the slogan and very rarely bought a non-Intel PC as a consequence.)

But I believe that this slogan will be rapidly replaced by “AI inside” because I believe we are almost at the point when ALL future apps will include elements of AI. I also believe there is a very good chance that Amazon’s Alexa might become the de facto automatic speech recognition platform that will sit in front of (outside) every single app in the future.  (My rationale is here.)

Why do I say this?

First, you need to recognize that AI is not one singular, all-embracing technology. Rather, it is a set of technologies that hope to emulate the way a human interprets and acts upon information — albeit at the speed of light and (we hope) without error, on a 24/7 basis. As such, AI includes technologies such as natural language (voice) processing (NLP), semantic analysis and cognitive processing.

Second, these technologies have become pervasive. The CEO of IBM recently announced at Davos that Watson (a supercomputer and a collection of AI APIs) is now having a (positive) impact on the lives of some billion people (about 1/7th of the world’s  population). I don’t know how many Echo and Dot units have been sold by Amazon (it must be tens of millions, at least) but each unit gives you access to Alexa, which uses both voice recognition and processing.

See also: 10 Questions That Reveal AI’s Limits  

Third — and most important — you don’t need to have a degree in AI (any more) to deploy AI.

AI was notionally conceived by Alan Turing in 1936 (but, in one sense, you can trace the origins of AI all the way back to Archimedes!). I was taught elements of AI at university in the early 1970s, but I didn’t have a chance to develop an AI app until the mid-1990s when I was a consultant for the Nationwide Building Society. We had just finished a ground-breaking piece of work that involved the development and deployment of the world’s first touch-screen-driven, customer self-service system. This system was a huge success on all measures, so the client I was working for at the time asked me:

“How far can you take this idea? Could you, for example, develop a system that’s as good as — or even better than — our best sales person?”

Without knowing it at the time, he was asking me to develop our first AI system. Fortunately for me (because I am certainly not an AI expert), Accenture had just hired an authority on the subject. He was swiftly assigned to my project team, and we stepped once more into the unknown world of innovation.

We started by gathering a team of the client’s top sales people. We then sat them down with our AI expert, who had been carefully briefed on the rules governing the sale of regulated products. We also called in the services of a user experience designer to obtain a better understanding of people’s risk appetites and option requirements. Last, but certainly not least, we asked a group of customers to help us develop the system that would be designed for their stand-alone use.

The result blew everyone away.

It even won the support of the U.K. Financial Services Authority (FSA), which agreed to assess the system for compliance. The FSA tested and analyzed every aspect of the new application — and then signed off. It was the first time the FSA had ever approved a sales platform that removed the need for a sales person.

Remember, this happened in the early ’90s — long before Java, Windows 95 and the first PlayStation were launched. Our system is a tribute to a client who not only had the vision to see the possibilities but also had the courage to take on the challenge — as well as the very real risk of failure.

However, there is a sad but rather revealing postscript to this story.

What happened to this ground-breaking system? Well, it was lauded, feted and widely acclaimed — and then quietly shelved. The building society decided to focus on building its Systems of Record (SoR) rather than its Systems of Engagement (SoE). And, sad to say, that was not an uncommon fate back then. Real innovation is often too radical for most risk-averse management to stomach. Sometimes it takes time to build an appetite for the truly ground-breaking. And maybe — just maybe — 20 years later, that time has come.

There was another problem: I only had one AI programmer at my disposal, and there weren’t that many more in the U.K. at the time. Given this, it would have taken a considerable amount of time to build an industrial-strength application that could have been put into the hands of any customer. But now we don’t have that problem.

One of the firms we at Clustre represent is an AI consultancy that is AI-technology-agnostic. It conceives, designs and builds AI-driven customer and employee apps that use a variety of AI technologies — as appropriate. It was recently asked by a loyalty card operator to show how AI could be used to allow a card holder to get an answer to a query without talking to a human or having to scour through FAQs (which I think are generally pretty useless). The firm created a web-based chat bot that used Watson to help recognize and understand the question and used another product to drive the Q&A process and, ultimately, answer the question.

So clever is the bot that it can easily handle misspellings and allow the questions to be phrased in a variety of ways and still operate properly. I would hazard a guess that this tool could handle at least 50% of all customer queries (the rest would be handed off to a human to resolve). That’s a lot fewer calls that need to be routed through to a human.

See also: Why 2017 Is the Year of the Bot  

So, you may ask, how many days did it take our AI consultancy to design and build this AI-driven chat bot? Just five.

Five days to design and build a tool that could potentially reduce call center volumes by around 50%!!!

AI has truly arrived, and everyone should be looking at how you are going to deploy it NOW!

fat tax

Should You Announce How Fat Workers Are?

A shockingly serious proposal has been floated to first persuade (and later possibly compel) publicly traded companies to disclose to shareholders quite literally how fat their employees are.

Also, how much they drink, how well they sleep and how stressed and depressed they are.

This proposal, advocating what is known as a fat tax, shouldn’t even merit a discussion among rational businesspeople, and yet here we are, discussing it. Even Harvard Business Review (HBR) is discussing this.

Why? Because the well-financed, well-organized cabal behind this fat tax proposal include corporate names like Johnson & Johnson, PepsiCo, Humana, Merck, Novo-Nordisk and Unilever. The leader of this group is a South African insurer called Discovery Health.

If you guessed that any critique written by me would also implicate Ron Goetzel, you would be correct. Despite having now himself admitted that most wellness programs fail, he is the one justifying this entire scheme by claiming that wellness programs increase stock prices — even though they don’t. We’ve already offered a completely transparent analysis to the contrary.

He also made a rookie mistake in his own analysis. The stock prices of companies in his study diverged greatly in both directions from the averages, and he didn’t rebalance existing holdings annually. It’s simple compounding arithmetic. Suppose the stock market rises X% a year. If every stock in your portfolio increases at that rate, you’ll match the averages. However, if half your stocks increase 2X% a year while the other half don’t appreciate at all, and you don’t rebalance, you’ll beat the averages. Simply by doing nothing.

Goetzel’s study appeared right before the fat tax proposal was floated at Davos. No coincidence here — Discovery Health (the sponsor of the Vitality Institute) cites the study as a basis for wanting shareholders to “pressure” companies into disclosing the number of fat employees they have. And the more fat employees a company has, the more shareholders will insist on wellness programs, thanks to this study. Johnson & Johnson and Discovery both sell wellness programs, while Merck and Novo-Nordisk sell drugs for various wellness-related conditions.

We urge reading the HBR link in its entirety to see why a fat tax would be even worse than it sounds. Some highlights:

Most importantly, though – and you don’t need Harvard to learn this – it’s just not nice to stigmatize employees for their weight or other shortcomings unrelated to job performance. Basic human decency should have been taught to this cabal a long time ago.

We’ve pointed out many times in ITL that these wellness people were absent the day the fifth-grade teacher covered arithmetic. This proposal suggests that they were also absent the day the kindergarten teacher taught manners.

How Google Is Wrong About the Internet

Eric Schmidt, the executive chairman of Google, said this week that the Internet will disappear — “There will be so many IP addresses, so many devices, sensors, things that you are wearing, things that you are interacting with, that you won’t even sense it,” he said. Now, Eric is a very smart fellow; he’s worth several billion dollars more than I am (the score is Schmidt, $8.3 billion, me, $0 billion); and he even has a better hairline than I do despite being two years older. He made his comments in Davos at the World Economic Forum, known as the gathering spot for very serious people. So his remarks have been getting quite a bit of attention and consideration. 

But he’s wrong.

He’s wrong for the same reason that people have been wrong since I started covering technology for the Wall Street Journal going on 30 years ago. That suggests to me that people will keep being wrong for the same reasons for some time to come, including in the world of insurance, where we are all having to try to figure out how the Internet of Things will play out. So let me point out the two issues that mean that even very smart people in very serious settings can’t just assume the sort of technological utopia that Schmidt is describing.

They are:

  • Decision rights
  • Transaction costs

Let’s look at those issues in the context of an article in the New York Times many years ago that got me mad enough to start thinking about the blind spot in the first place. A very bright reporter, and something of a friend, began by painting an idyllic vision of an automated future: A person hopping out of bed would step on a sensing device that would let the house know he was up. The house would then turn on CNN in the family room, start the coffee and probably do some other things that I no longer remember at this remove.

Nice image, right?

Decision rights

But what if I don’t want to watch CNN? What if I’m more interested in watching ESPN that morning? Or my kids were already awake and watching cartoons — would my stepping on the pad change the channel despite the screams that would surely result? What if I’m heading off to meet someone for breakfast and will have coffee there, not at home that day?

A key question with any sort of automation is: Who owns the decision rights? In the case of CNN and the coffee, do I want the house to have the decision rights, or do I want to retain them?

Transaction costs

How much effort do I have to put into the automation? Is it really worth it to put a sensor under my carpet or even to lay something on top of the carpet? What does that cost? How long does it take me to configure the TV and other systems in the house so that they react appropriately?

Those transaction costs then have to be compared against the benefits, which, in the case of CNN and the coffee, are trivial. It’s just not that hard to pick up the remote and click the TV on or to fix the coffee in the morning (which you would have had to do before going to bed in the automated scenario.)

People tend to get so excited about the George Jetson-like possibilities that they ignore decision rights and transaction costs and paint visions that simply won’t occur in any reasonable timeframe.

That’s how we ended up with:

— The talk back in the early ’90s about “agents” that would pull together what some called “The Daily Me,” a personalized newspaper that would gather all the news that it knew you were interested in and mix it in with your schedule and other things to lay out your day for you. The problem was that these personalized papers took a huge amount of effort and were so inaccurate that no one would turn all the decision rights over to an agent. What if you didn’t have time to read the news that day? What if you had become interested in some topic that you’d never read about before — how would your agent know?

(I took the talk of agents somewhat personally because the three pieces I wrote for the Wall Street Journal that easily got the most response from readers in my 17 years there were about: a time I sailed across the Atlantic in a small boat, having never sailed before, in what turned out to be some monster storms; my two-week career as a professional wrestler; and my mother (a piece written with my younger brother). I guarantee you that no one who picked up the Wall Street Journal the day those pieces appeared was looking for anything about sailing, professional wrestling, my mother or me, so no one would have ever seen them in a world of agents.)

This talk of agents is cropping up again, by the way, and is surely part of the reason that Schmidt wants to talk about having the Internet disappear. Google wants to make search so efficient that its engine knows what you want to find even before you think to look. The company has made impressive strides — if you type in Elm Street while looking for directions on your Android phone, Google Maps usually guesses quickly and correctly which Elm Street you want — but that’s a long way from the world that Google is describing, and the transaction-cost and decision-rights issues will still get in the way.

— The fuss over the “Internet refrigerator” that still crops up from time to time. The idea is that your refrigerator would sense when, say, you were low on milk and reorder it for you. But that requires an awful lot of engineering, both in the refrigerator and in whatever system of grocery delivery would be used, and only makes sense if you’re turning just about all your shopping over to your refrigerator — if you have to go to store anyway, it’s simple to grab some milk.

And there is always the issue of decision rights. What if a family goes on vacation? How long will the refrigerator keep ordering? I have a friend whose 21-year-old son drinks a gallon of whole milk a day. When the son — who is 6’5″, weighs 285 pounds and looks like he could bench press a cow — is home, they can’t buy milk fast enough, but when he’s gone at school they don’t need any. Do they have to let a few gallons of milk sour before the refrigerator figures out the son is gone?

— The excitement about home controls: remote-controlled lighting, the Internet thermostat that will sense who’s in a room and adjust lighting levels and temperature to personal preferences and so on. Those are just an awful lot of work for not much benefit — you can always flip a light switch or adjust a rheostat — and doesn’t resolve the issues that come up when one person likes a room cooler than the other.

Technology will make plenty of tasks disappear, but let’s not be too hasty. We need to think through the costs, the benefits and the potential for errors and conflicts in automated systems, to make sure we don’t fall victim to the seductive tendency to ignore transaction costs and decision rights.

The Internet won’t disappear in my lifetime, which I’m assuming will be at least 30 more years. I won’t even have sensors that turn on CNN and start my coffee when I get out of bed.